Defender's Corner

A year-long study by security awareness platform Pistachio sent 2.47 million simulated phishing emails to 123,000 employees across 1,200 organisations, and found that click rate — the metric most phishing training programmes are judged on — tells only part of the story, according to SecurityWeek.

What actually matters is what happens after the click: 30% of tech and IT workers clicked at least once, and nearly 20% of construction employees handed over credentials after clicking. A click with no credential submission creates no real breach.

What you should do: If your organisation runs phishing simulations, push for reporting that tracks credential submission and suspicious-email reporting rates alongside clicks. If you're an individual, the habit worth building is simple: before entering a password anywhere, check the URL bar first.

An API key (a credential that lets one piece of software authenticate with another) left in a public code repository is a gift to attackers. Bots scan public repositories constantly, and the majority of exposed keys remain active long after they are found.

A security researcher has proposed an open standard called ORKS (Open Revocable Key Standard), detailed at Security Week, that would allow anyone who finds a leaked key to destroy it immediately by submitting it to a revocation endpoint — no account or authentication required.

The standard is not yet widely adopted, but the underlying lesson is actionable right now: if your team uses API keys, set a reminder to rotate them on a schedule and audit whether any have been committed to public repositories using a tool like GitHub Secret Scanning.

The UK's National Cyber Security Centre (NCSC) has published a guide on shadow AI — the use of AI tools that employees adopt without their employer's knowledge or approval, according to the NCSC blog.

One survey cited in the piece found that 71% of employees use AI tools their employer has not approved. The risk is straightforward: when you paste work data into a consumer AI service, that data may be stored, retained, or used to train the service — outside any security controls your organisation has in place.

The NCSC's advice is practical: before using any AI tool for a work task, ask whether it is on your employer's approved list. If you are not sure, ask. If no approved option exists for your needs, raise it with your IT or security team rather than improvising. The goal is not to stop using AI — it is to use it without handing your company's sensitive data to an unknown third party.

HaveIBeenPwned is a free service that tells you whether your email address or phone number has appeared in a known data breach. It is run by respected security researcher Troy Hunt and is widely trusted.

With 8.8 million records from the Manchester Airports Group breach now in its database, this is a good moment to check. Go to haveibeenpwned.com, enter your email address, and see which breaches it appears in. You can also set up free alerts so you are notified automatically if your details appear in future leaks.

If your email is listed, update passwords on affected accounts and switch on two-factor authentication (a second verification step beyond your password) wherever it is offered.

Anthropic has warned Claude users that infostealer malware — software that quietly copies saved passwords, browser cookies, and login credentials — was used to hijack accounts and drain usage limits, according to SecurityWeek. The malware arrives via unofficial downloads or fake apps, not through Claude itself.

If your Claude usage limits appeared to drain without you doing anything, that is a warning sign. Anthropic has signed out affected sessions and removed saved payment methods as a precaution.

The practical step: run a full scan with Malwarebytes — a free, widely trusted tool — before re-adding any payment details to Claude or any other service. Then change your passwords for any accounts that were logged in on the affected device. This applies whether or not you use Claude: infostealers target every saved credential on your machine.

The UK's National Cyber Security Centre has issued a warning about increased attacks on operational technology (OT) systems — the industrial software and hardware that runs power grids, water treatment, and manufacturing, according to the NCSC advisory.

Their first recommendation applies to everyone, not just industrial operators: know what you have exposed to the internet.

For home and small business users, the practical version is this: log into your router's admin panel and check which devices are reachable from outside your network. Most routers have a "port forwarding" or "exposed services" section. If you see something you did not intentionally open up — a printer, a camera, a server — close it.

If you manage any internet-connected device for work, change its default password today. Default credentials are the first thing attackers try, and they work more often than anyone should be comfortable admitting.

WhatsApp has rolled out several security improvements worth enabling now, according to Security Week.

Two-step verification (2SV — a second layer of protection beyond your SIM card) has been upgraded from a six-digit PIN to a full password that can include letters, numbers, and special characters. Longer and more complex is better. Go to Settings → Account → Two-step verification and update yours.

Passkeys (a login method that replaces passwords with a cryptographic key stored on your device) can now be added in multiples, which is useful if you switch between iOS and Android. Android users also get a new caller context feature that shows the country and any shared group membership for unknown callers — a quiet but useful tool against number-spoofing scammers.

The Uber fine (covered below) is a useful reminder that EU law gives individuals the right to challenge automated decisions — and similar rights exist in the UK, Canada and California under their respective data protection laws.

If a company tells you that an automated system has denied you a service, suspended an account, or made a significant decision about you, you can ask for a human review. In the EU and UK, this right is explicit under GDPR and UK GDPR. To exercise it, contact the company in writing, reference "automated decision-making," and request that a person reviews the outcome.

It will not always work quickly. But companies that ignore such requests face exactly the kind of regulatory attention Uber just received. Know the right exists — it is worth using.

Anthropic — the company behind the Claude family of AI models — has expanded a programme called Project Glasswing, which gives security teams access to its most capable AI model, Mythos 5, according to Security Week. It has also launched a $35 million fund to help open source projects find and fix vulnerabilities.

For most readers, the practical takeaway is this: Claude Security, available now in public beta for Claude Enterprise users, can scan your codebase, flag vulnerabilities with severity ratings, and suggest fixes — with a human required to approve anything before it deploys.

If your organisation uses Claude Enterprise, check whether Claude Security is enabled for your team. If you maintain open source software, the Defender Advantage Fund (0xDAF) is worth looking into for grant support.

Email filters were built to catch known threats. AI-generated phishing emails are designed to look like normal business correspondence, personalised using public information from LinkedIn and company websites. A Harvard Business Review study found AI-crafted spear phishing (targeted, personalised phishing) hit a 54% click-through rate. Filters alone will not cut it.

The article from Bleeping Computer makes a useful point: assume some phishing will always reach the inbox. The better question is what you have in place once it does.

What you should do: Pair your email filter with endpoint detection software and identity monitoring. Enable multi-factor authentication on every account so that a stolen password alone cannot open the door. When in doubt about an email, call the sender directly using a number you already have.

Microsoft has removed WMIC (Windows Management Instrumentation Command-line) from Windows 11 versions 24H2 and 25H2, according to Bleeping Computer. WMIC is a built-in Windows utility that attackers have long repurposed to delete backups, disable antivirus software, and dig deeper into compromised systems — all without triggering alerts, because the tool was signed by Microsoft itself.

Its removal means a whole category of well-established attack techniques simply stops working on updated systems.

What to do: Open Windows Update and make sure you are running the latest version of Windows 11. If you manage Windows machines for a small business, check that automatic updates are enabled across all devices. This one costs nothing and closes real doors.

GitHub's Dependabot is a built-in tool that automatically scans your code's dependencies (the third-party packages your project relies on) and alerts you when something dangerous turns up.

Until this month, Dependabot's malware alerts only covered npm, one of the most popular package registries. That meant poisoned packages from PyPI, Maven, RubyGems, NuGet, Go, crates.io, and PHP Composer went undetected. As reported by Help Net Security, all eight ecosystems are now covered.

If you manage any GitHub repository, enable Dependabot malware alerts now. Go to your repository, click Settings → Security & analysis → Dependabot alerts, and switch it on. It is free, takes thirty seconds, and runs across more than 30 million repositories already.

The macOS Screen Sharing vulnerability being actively exploited right now is a good reminder that built-in remote access features are a liability when left on by default. Screen Sharing, Apple's remote desktop tool, only needs to be reachable from the internet to become a target — and many users have no idea it's enabled.

Disabling it takes thirty seconds. On your Mac, open System Settings → General → Sharing, find Screen Sharing, and switch it off. If you do need remote access to your Mac, consider using it only over a VPN rather than exposing port 5900 directly to the internet.

More broadly, go through everything listed under Sharing in System Settings. Remote Login, Remote Management, and AirPlay Receiver are all features worth disabling if you don't actively use them.

WhatsApp is rolling out a new feature called Scam Alert, according to Bleeping Computer. It runs a machine learning model directly on your device to flag messages from unknown contacts that match known scam patterns.

Nothing leaves your phone for analysis. No message content is sent to WhatsApp, Meta, or anyone else. If the model spots something suspicious, you will see a prompt to block, report, or continue. If it gets it wrong, you can mark the chat as trusted.

The feature is in limited beta right now but is expected to roll out more broadly. To check if it is available: open WhatsApp, go to Settings → Privacy and look for Scam Alert. Enable it if you see it.

Scam messages via messaging apps are one of the most common ways people are defrauded. This is a free, privacy-respecting layer of protection that takes thirty seconds to enable.

Signal has launched Automatic Key Verification, a new feature described by Bleeping Computer that protects against man-in-the-middle attacks (where someone intercepts your messages by quietly swapping out your contact's encryption key for their own). Previously, confirming a contact's key required meeting in person or using a separate channel. Now Signal uses Cloudflare and Trail of Bits as independent auditors to verify that the key associated with a phone number has not been tampered with.

To enable it: open Signal, go to Settings > Privacy > Advanced, and toggle on Automatic Key Verification. A green "Encryption verified" badge will appear when a chat is confirmed clean.

This matters because Russian state-sponsored hackers previously compromised Signal accounts by abusing the Linked Devices feature. This feature closes a gap they exploited.

Google announced that Chrome's anti-abuse systems blocked over 7 billion unwanted Android notifications per day in Q1 2026, according to Bleeping Computer.

Notification abuse is a real attack channel: scammers use browser notifications to deliver phishing links and fraudulent payment requests. Chrome now automatically revokes notification permissions from inactive or suspicious sites and lets you review what it has removed.

What to do right now: On Android, open Chrome and go to Settings > Notifications. On desktop, go to Settings > Privacy and security > Site Settings > Notifications. Review which sites have permission to send you notifications and revoke anything you do not recognise. Chrome's Safety Hub will also show you permissions it has already revoked automatically — check that list too. If a site you removed turns out to be legitimate, you can restore it.

This takes about two minutes and cuts a surprisingly effective scam delivery channel off at the source.

If you or your team write Python code, your encryption just got a significant upgrade available with a single install. The pyca/cryptography library now includes support for ML-KEM and ML-DSA, the two post-quantum cryptography standards finalised by NIST (the US National Institute of Standards and Technology). Post-quantum cryptography refers to encryption algorithms designed to resist attacks from quantum computers, which are expected to eventually break most encryption in use today.

You do not need a quantum computer bearing down on you to act. The point of switching now, as the project puts it, is precisely that there is no emergency yet — which means time to migrate carefully rather than frantically.

To get it: pip install cryptography (version 44 or later).

If you are not a developer, ask whoever manages your software stack whether your tools are on the roadmap for post-quantum upgrades.

GitHub's Dependabot is a built-in tool that automatically scans your project's dependencies — the third-party packages your code relies on — and alerts you when something is outdated or dangerous. Until now, its malware detection only covered npm (a package registry for JavaScript developers). That changed this month, according to Help Net Security.

Dependabot now draws from OpenSSF's malicious-packages database, a public feed covering eight ecosystems including PyPI (Python), Maven (Java), RubyGems, NuGet (.NET), Go, crates.io (Rust), and PHP Composer.

If you use GitHub for any development work, enabling Dependabot malware alerts takes one minute and adds an automatic watchdog across your entire dependency chain.

What you should do: Go to your GitHub repository or organisation settings, find the Security section, and enable Dependabot malware alerts. It is free and takes effect immediately.

The RovoBlast finding, presented at DEF CON 34, illustrates a risk that applies well beyond Atlassian: AI assistants with broad integration access can be weaponised to exfiltrate data automatically once an attacker controls their input.

Varonis researchers recommend a concrete set of steps that any team using an enterprise AI tool can apply right now, according to SecurityWeek:

1. Audit which systems your AI assistant can access and remove any integrations not actively used.
2. Block AI access to sensitive departments — HR, legal, and finance — unless there is a specific, reviewed business need.
3. Disable autonomous browsing or multi-step agent features if your team does not rely on them daily.
4. Enable activity logging for your AI assistant and review it regularly for unusual data access patterns.

Think of it like giving a new intern a master keycard on day one — useful in theory, but you would probably hand them the specific keys they actually need instead.

The Swiss government breach is a useful reminder: it is not just about patching. It is about knowing what is exposed in the first place.

Microsoft SharePoint — the document storage and collaboration platform used by thousands of organisations — often ends up with external internet access enabled by default, even when that access is not needed. That is unnecessary attack surface.

Do this today:

1. Open your Microsoft 365 admin centre and check SharePoint's external sharing settings under Policies → Sharing
2. If external sharing is not required for your team, set it to "Only people in your organisation"
3. Confirm your July 2026 SharePoint patches are applied via Microsoft 365 admin → Health → Message Centre

Takes under 10 minutes and meaningfully reduces your exposure.

SharePoint admin sharing settings — Microsoft Docs

Microsoft has updated its free Zero Trust Assessment tool to include a dedicated AI security pillar, according to Help Net Security.

Zero trust (a security model built on the principle of "never trust, always verify" — no user or system is trusted by default, even inside your own network) is increasingly relevant as organisations deploy AI agents with access to sensitive tools and data.

The tool automatically scans your Microsoft security configuration, identifies gaps, and now produces a prioritised roadmap covering identity, devices, data, network, and AI systems. It is free for Microsoft environment users.

If your organisation uses Microsoft 365 or Azure, run the assessment at Microsoft Security. Pay particular attention to the AI pillar if you have deployed any Copilot or AI agent tools. The results will tell you exactly what to fix first.

Cloudflare has announced AI agent wallets with built-in spending controls, according to Help Net Security. The feature lets you create a Virtual Wallet for any AI agent you run, with hard caps on how much it can spend, which services it can pay for, and per-transaction limits — all controlled by you through an Account Wallet.

This directly addresses a real risk: AI agents given open access to APIs and payment methods can rack up unexpected charges or be manipulated into spending beyond their intended scope.

Handle reservations are open now at Cloudflare. If you use AI agents for any work task, reserve a handle and review what spending permissions your current agents already have. Giving an AI agent uncapped access to APIs or services is roughly equivalent to handing a new employee an unlimited corporate card without a receipts policy.

The full announcement is at helpnetsecurity.com.

NuGet is Microsoft's software package repository for .NET developers — think of it as an app store for code libraries. Starting 17 August 2026, new API keys (passwords that allow developers to publish software packages) will expire after 30 days instead of 365, according to Help Net Security.

Why does this matter? A stolen long-lived key is a skeleton key that stays valid for a year. A recent attack used stolen credentials to push a malicious package that was downloaded 6,000 times in 36 minutes. Shorter lifetimes cap the damage window.

If you are a developer publishing to NuGet.org, audit every workflow that uses an API key and replace any key created before 17 August before 1 November 2026. Better still, migrate to NuGet Trusted Publishing — it uses short-lived, automatically expiring credentials instead of permanent keys.

Malware on Windows and macOS can quietly write fake enterprise policy keys to your system, then use those keys to force-install browser extensions that replace your search engine or new tab page — and Chrome treats them as administrator-approved, making them impossible to remove. Google is preparing a fix, as reported by Bleeping Computer.

The feature is not live in stable Chrome yet, but when it ships, it will automatically block these fake policy installs on personal devices and uninstall hijacking extensions that were previously locked in.

What to do right now: If your Chrome shows "Managed by your organisation" and you do not work on a company device, that is a red flag. Type chrome://management in your address bar. If it says your browser is managed but you do not recognise the organisation, run a malware scan with Malwarebytes Free. Keep Chrome updated so the new protection arrives as soon as Google ships it.

A live AMOS (Atomic macOS) stealer campaign is targeting Mac users through fake "macOS toolkit" websites. The trick is simple: a site presents copy-paste instructions that look like a legitimate software setup step. You open Terminal (Mac's command-line tool), paste the text, and unknowingly install a credential-stealing app that harvests passwords, browser data, and cryptocurrency wallet files.

According to SANS ISC, the malware phones home to a remote server and begins collecting data immediately.

The fix is equally simple: never paste anything into Terminal from a website unless you have read it character by character and understand what it does. Legitimate software installers do not require this. If a site tells you to open Terminal and paste something, close the tab.

If you have already done this, change your passwords and check your accounts for unusual activity today.

CISA issued an urgent alert after attackers disrupted more than 30 community water systems in Minnesota by targeting internet-exposed PLCs (programmable logic controllers — the industrial computers that manage physical equipment like pumps and valves), according to Bleeping Computer.

Attackers changed passwords to lock operators out, modified IP addresses to disconnect devices, and forced some utilities to switch to manual operations.

CISA's primary instruction is simple: remove these devices from direct public internet exposure immediately. If that is not possible, place them behind a VPN (a secure, encrypted access tunnel) or gateway device. Change all default passwords and restrict access to a known list of approved IP addresses.

This matters beyond utilities. The same principle applies to any remotely accessible device you manage — routers, smart home hubs, or office equipment. Anything reachable from the public internet without a barrier is a target.

What you can do today: Check whether any devices in your home or workplace are directly accessible from the internet. Your router's admin interface is a good place to start.

Most security advice focuses on keeping attackers out. A real incident documented by Huntress, covered on Bleeping Computer, shows why what happens next matters just as much.

In the case Huntress investigated, the attacker gained entry through a SQL injection flaw (a vulnerability where a website passes user-typed input directly into a database without checking whether it is safe), then quietly created a new admin account, enabled remote desktop access, disabled Windows Defender, and installed malware on the web server — all before anyone noticed.

The takeaway for non-technical readers: attackers rarely rush. They settle in.

What to do: Check your active user accounts periodically and remove any you do not recognise. If you manage a website, ask your hosting provider whether input validation and web application firewall protections are in place. Huntress offers free resources for small businesses at huntress.com.

Research from Vanson Bourne surveying 600 senior IT security professionals found that 73% of organisations admit they would not be fully prepared if a serious cyberattack happened tomorrow, according to The Hacker News. The biggest gap is not tools — it is coordination. Legal, communications, and executive teams often arrive late when speed matters most.

The fix is a tabletop exercise: a scripted, hypothetical attack scenario that your team walks through together without any real systems at risk. Think of it as a fire drill for a breach.

You do not need a security consultant to start. CISA publishes free tabletop exercise packages at cisa.gov/resources-tools/services/cisa-tabletop-exercise-packages. Pick a scenario that fits your organisation, schedule two hours, and get legal and leadership in the room. The goal is to find the gaps before an attacker does.

CISA, the FBI, and their Australian counterparts have published free guidance called CI Fortify — Advice for Isolating Vital Systems, aimed at helping organisations disconnect critical systems from their networks during an attack while keeping essential services running. Read it here via Bleeping Computer.

The core exercise it recommends translates directly to smaller operations and even households: map every connection your most important systems have to the outside world, and decide in advance exactly where you would cut them if something went wrong.

For a small business, that means knowing which systems matter most, which vendors and cloud services they depend on, and how long you could operate manually if those connections were severed. For individuals, it is worth knowing where your router admin panel lives, what your ISP support number is, and whether your most critical files exist somewhere offline.

Knowing the answer before an incident is the difference between a plan and a panic.

A supply chain attack (where hackers poison a software package that developers trust and use, sending malicious code downstream to every project that installs it) is one of the most effective ways to compromise thousands of organisations at once.

GitHub and PyPI — the Python Package Index, the main repository where Python software packages are published — have both introduced new protections, according to Security Week.

GitHub's automated dependency tool, Dependabot, now waits three days before updating your project to a brand-new release. That gap gives security researchers time to catch a poisoned version before it reaches your codebase. PyPI is now blocking new files from being added to releases older than 14 days, closing a gap where an attacker with a stolen publishing token could quietly poison a long-trusted package version.

What to do: If your team uses Dependabot, check that your dependabot.yml configuration is not overriding the new cooldown behaviour. If you maintain Python packages on PyPI, no action is needed — the protection applies automatically.

CTM360 researchers documented an evolved insurance phishing campaign, reported by The Hacker News, where attackers no longer just collect your credentials to use later. Instead, a fake insurance portal acts as a live middleman: as you type your username and password, the attacker simultaneously submits them to the real insurance site in real time, completing the login before you realise anything is wrong.

The primary entry point was fake Google Ads promoting insurance quotes and renewals.

CTM360's report does not identify a single technical control that defeats this entirely, but the most reliable protection it implies is avoiding the attack's starting point. Before entering any credentials on an insurance or financial portal, type the organisation's URL directly into your browser rather than clicking any advertisement or search result link. Check the full URL in your address bar carefully before submitting anything.

Meta has launched Facebook Verified, a free identity badge for personal accounts, according to Help Net Security. To get it, you record a short video selfie. Facebook checks it against your existing profile photos to confirm you're a real person.

The badge appears on your profile and in Marketplace, Dating, and Groups — exactly the places where AI-generated fake accounts are most dangerous. It costs nothing and takes a few minutes.

If you use Facebook Marketplace or Dating, this is worth doing now. A verified badge signals to other users that you passed an identity check — and it makes your account a less attractive impersonation target.

To get started: go to your Facebook profile settings and look for the Verified badge option. You must be 18 or older with an account in good standing.

Microsoft is changing how Windows enterprise activation works, replacing a software-only trust model with hardware-backed verification using TPM (Trusted Platform Module — a dedicated security chip built into most modern computers that stores cryptographic keys and confirms the machine hasn't been tampered with), according to Help Net Security.

The change affects Windows Key Management Service (KMS), the on-premises tool organisations use to activate Windows across large fleets of machines. From August 2026, Windows Server 2025 will flag whether your KMS host is ready for hardware-based security. Mandatory enforcement arrives with the next Windows Server LTSC release.

What to do now: If you manage Windows servers, identify your KMS hosts and confirm whether the physical machines have TPM support. If they don't, hardware upgrades may be required before enforcement kicks in. Loop in your IT team and review Microsoft's rollout schedule ahead of the August readiness messaging.

PyPI (the Python Package Index) is the central repository where developers download Python software libraries. As of this month, it now rejects any new files uploaded to a release older than 14 days.

This matters for non-developers too. If an attacker compromises a developer's credentials, one classic move is to slip malicious code into an older, trusted package version. Users downloading what they believe is a stable, long-tested release would get the poisoned version instead.

The new rule shuts that window. Older releases are now frozen. Attackers cannot retroactively tamper with them even if they steal the right credentials.

You do not need to do anything to benefit from this change. It is already live for all packages on PyPI.

What to do: If your organisation's software teams use Python packages, ask them to confirm they are pulling dependencies from PyPI directly rather than caching old versions locally from before July 2026.

ServiceNow is enterprise software used by organisations to manage IT helpdesk tickets, HR requests, and internal knowledge bases. Security firm Bishop Fox found that 31% of the 166 ServiceNow instances they tested during authorised engagements returned records or confirmed record counts to completely unauthenticated requests, according to Help Net Security.

Bishop Fox released Snowpick, a free open-source scanning tool that checks your own ServiceNow instance for this exact misconfiguration. It works by grabbing the public session token your portal issues to anyone who visits the login page, then using it to probe default widgets and API endpoints for accessible data.

If you manage a ServiceNow instance, run Snowpick against it before someone else does. The tool is available on Bishop Fox's GitHub. If you are an employee, ask your IT team whether your organisation's ServiceNow portal has been audited for public data exposure.

Amazon GuardDuty — a managed threat detection service that monitors AWS cloud accounts for suspicious activity — has launched an AI-powered investigation agent, now in public preview and free to use during the preview period, according to Help Net Security.

When a security alert fires, the investigation agent automatically analyses the finding, maps it to the MITRE ATT&CK framework (a standardised catalogue of attacker techniques), assesses risk and confidence levels, and produces actionable remediation steps. Think of it as the first analyst on the scene writing up the incident report before your team even opens their laptop.

It is available in 10 AWS regions, limited to 10 investigations per account per day during preview.

What you should do: If your organisation runs workloads on AWS and has GuardDuty enabled, turn on the investigation agent in your GuardDuty console today. It costs nothing during preview and reduces the window between alert and action.

Microsoft has released Dusseldorf, a free open-source platform for out-of-band application security testing (OAST) — a method of catching vulnerabilities that don't leave obvious traces inside an application but cause the app to quietly make unexpected outbound network connections.

It detects flaws like SSRF (server-side request forgery, where an attacker tricks a server into making requests on their behalf), cross-site scripting, and XML external entity injection — all by watching what the application reaches out to at the network edge.

Dusseldorf runs DNS, HTTP, and HTTPS listeners on a domain you control. Every unexpected request gets logged. It is designed for private deployment and comes with a graphical interface and API.

It is free on GitHub and built for both local lab setups and Azure cloud environments.

Get it: Microsoft Dusseldorf on Help Net Security

The 2026 SANS AI Survey found that 78% of security practitioners now use generative AI in their daily work — but the same research, covered by Help Net Security, highlights that human judgement remains the most effective check against AI-driven attacks.

Separately, researchers at Tracebit found that planting "context bombs" (deliberately misleading information seeded into environments to confuse AI agents) significantly slowed automated attackers from completing their objectives.

You do not need a security team to apply the core principle: slow down and question anything that creates urgency. Phishing and social engineering attacks, including the ClickFix technique spreading ACR Stealer, work because they pressure you to act before you think.

What to do: Before clicking, pasting, or downloading anything unexpected, pause for ten seconds and ask: did I request this? If the answer is no, close the window.

Apple has published a warning that scammers are using FaceTime to impersonate banks and Apple itself, per Help Net Security.

The playbook goes like this: a caller spoofs (fakes) a trusted number, drops some personal details early to sound legitimate, then creates urgency so you stay on the line. The goal is to get you to share your password, a two-factor authentication (2FA) code, or disable security features entirely. Some callers instruct victims to turn off 2FA or Stolen Device Protection, claiming it will stop an attack. It will not. It opens the door to one.

Apple is clear: it will never ask you to share a password, a 2FA code, or to accept a 2FA prompt over the phone.

What to do: Hang up. Call the organisation back on a number you look up yourself. If you receive a suspicious FaceTime call from someone claiming to be Apple or a bank, take a screenshot and send it to reportfacetimefraud@apple.com.

1Password has released a beta integration with Claude that lets the AI assistant complete browser tasks requiring a login — without ever seeing your actual password, according to Help Net Security.

Here is how it works: when Claude hits a sign-in page during a task, it requests the credential from 1Password. You approve the request using Touch ID or another biometric method. The password is filled in through a secure channel that never enters Claude's memory or Anthropic's systems. Once the task ends, the permission resets.

This is available to paid Claude subscribers (Pro, Max, Team, or Enterprise) on macOS with the 1Password desktop app and browser extension installed.

If you use Claude for browser-based tasks, this is worth enabling. It keeps your credentials out of AI context windows entirely — which is exactly where they should not be.